OpenAI

We're looking for engineers with GPU programming experience/knowledge to help write high performance kernels and improve the Triton compiler. As our ML understanding changes, we frequently need new kernels to try out promising ideas at scale. We also often want to co-design an operation for GPU performance and ML effectiveness, so we usually have significant scope for creativity on the kernel side. Ideal applicants will have GPU programming experience (any vendor), with attention to high performance. Compiler expertise. Good understanding of performance bottlenecks of modern processors. Parallel algorithm design expertise. C or C++ programming experience. Applicants must have experience with low level performance engineering with vector (SIMD) instructions- CPU OK (GPU of course also okay!). Basic understanding of undergrad math, particularly multivariable calculus, linear algebra. We primarily use Python and CUDA C++. We have very competitive salary and benefits, and an exciting work environment. Onsite in SF required 3 days a week, with a possible exception for extremely talented applicants. We can often sponsor US visas for international applicants. If you're interested, email me at rajan@ our primary domain. Depending on your background we may ask for an GPU-programming interview or a small take-home GPU-programming project. Then we'd have a few more general technical interviews followed by an offer.